Julia Richman lost nine months waiting for a building permit. Not because the plans were wrong, not because the zoning was contested, not because the engineering failed review. Nine months. She lost them because nobody at the permitting office had looked at the application yet, and it sat in a queue, untouched, while the interest on the construction loan kept compounding and nobody at the city could tell her when it would move.
"We lost nine months while we waited for someone to look at the permit," Richman told HousingWire in May. She is now the vice president of government relations at Clariti, the company Denver just paid $4.6 million to make sure that kind of delay never happens again. Whether you read that as poetic justice or regulatory capture depends on how cynical you feel about the building department.
In March 2026, the Denver City Council approved a five-year, $4.6 million contract with Clariti for its CivCheck platform, an AI pre-screening tool that scans uploaded permit documents, maps files to city requirements, and flags missing information before applications reach human reviewers. The initial purchase was $1.05 million, with the remainder covering support and expansion through 2031, and implementation began the following month.
In the same budget cycle, Denver's Community Planning and Development Department cut 59 budgeted positions, according to The Denver Gazette, bringing its headcount from 310 to 251.
Nobody called a press conference to connect those two numbers. Nobody needed to.
Sixteen Cents on the Dollar
Run the arithmetic, because it is not complicated.
An experienced urban planner in Denver earns a base salary of approximately $80,000 per year, per PayScale's 2026 data, and plan reviewers holding International Code Council certifications tend to earn slightly more, so once you add the standard government benefits multiplier of 1.35 for health insurance, retirement contributions, and payroll taxes, total compensation per position lands near $108,000 annually.
Fifty-nine positions at $108,000: $6.4 million per year in labor that left Denver's planning department, replaced by CivCheck's annual cost of $920,000.
That is 14 cents for every dollar of labor eliminated. Round up generously, use a lower salary estimate, call it 16 cents, and the math does not change. Denver replaced a significant portion of its human review capacity with software that costs roughly one-sixth as much, and it did so in a single budget action that generated two separate press releases, neither of which mentioned the other.
What CivCheck Actually Does
Credit where it is warranted: CivCheck is not an autonomous approval engine, and Clariti does not pretend otherwise, because what it actually does is pre-screen. A sophisticated form checker that reads uploaded documents, compares them against a city's specific submittal requirements, and generates a list of what is missing, incomplete, or inconsistent before the application enters the human review queue. Reviewers still make the final call on code compliance. The AI never says yes. It only says "not yet, and here is why."
That distinction matters legally. Denver configured CivCheck as a decision-support system, not a decision-making one, and it operates within parameters established by city departments, applying rule-based checks tied to local codes and generating traceable outputs that reviewers can evaluate before making their own determination. On paper, it is the least controversial possible application of AI in government: catching obvious mistakes before expensive professionals have to.
But look at the budget spreadsheet.
San Jose Found 75% of Its ADU Permits Were Incomplete
Denver is not alone in discovering that most of what overwhelms a permitting office is not complex code analysis. It is paperwork.
When Clariti ran a pilot with the City of San Jose examining accessory dwelling unit applications, approximately 75% of submissions were incomplete: missing documents, blank required fields, site plans that referenced a parcel number from a different property entirely. These are not judgment calls that require an ICC-certified plan reviewer with fifteen years of experience. They are clerical errors that could be caught by a checklist, which is exactly what AI turns out to be excellent at.
Honolulu deployed CivCheck and cut residential plan review times by 70%, according to Clariti's own data, with reviews that once took up to 90 minutes dropping to 20 or 30 minutes and total review time falling 64%. Dawn Takeuchi Apuna, an official with the City and County of Honolulu, described the improvement as giving staff "more consistency and confidence throughout the process."
Consistency. Confidence. Fewer staff. Notice which word does not appear in the testimonial.
At Least Twelve Cities and Counting
Denver, San Jose, Honolulu, Seattle, Vancouver, Calgary, and Toronto are all running CivCheck. Meanwhile, Baltimore, Corona (California), Louisville, Boston, Los Angeles, and Harris County, Texas, have partnered with competing AI permitting tools. A Route Fifty analysis in May 2026 placed AI-assisted permitting among the fastest-growing municipal technology categories in the country. A White House executive order and bipartisan congressional legislation have both put permitting reform on the federal agenda, giving cities political cover to move quickly.
What none of these cities has published is a full accounting of staff reductions that accompanied or followed AI adoption. Denver at least left the numbers visible in its budget documents, which is more than most will offer.
Who Benefits, and a Harder Question About Who Doesn't
If you are a homeowner submitting an ADU application for the first time and you forget to include the site drainage plan, CivCheck will catch it before you waste four weeks in a queue that will reject you anyway, and that is a genuine improvement over what came before. Inexperienced applicants, the ones Richman described as generating the lowest-quality submissions, stand to gain the most from automated intake screening. Their mistakes are the simplest to detect and the most expensive to process manually.
Builders and developers benefit differently, because professional applicants already know what to include in a submittal package and their bottleneck is not intake errors but queue depth. If AI diverts 50% of incomplete applications back to applicants before they enter the review queue, the professionals behind them in line move forward faster. Denver's own 180-day shot clock for permit decisions, backed by a $10,000 refund guarantee if the city misses the deadline, creates a financial incentive to thin the queue by any means available.
Cut 59 positions and deploy AI, and the shot clock becomes survivable on a smaller payroll, which is the logic that nobody in city government will say plainly but the budget says for them.
What This Doesn't Prove
Three honest caveats. None of them changes the bottom line.
First, the 59 cut positions may not all be plan reviewers, since Community Planning and Development departments include administrative staff, inspectors, policy analysts, and support roles, and some of those cuts may reflect attrition, retirement, or reorganization that predates the CivCheck contract by years. Denver has not published a breakdown by role.
Second, correlation is not causation, even when the timing is conspicuous. Denver has been working toward permitting reform for years, launched a dedicated Permitting Office in 2025, and had already cut single-family and duplex processing times by 45% since 2023, before CivCheck existed. The AI may be one initiative among several, not the driver.
Third, this analysis uses PayScale salary estimates rather than Denver's actual payroll data, and government compensation varies by classification, seniority, and bargaining unit. The 16-cent figure is an approximation, and the real ratio could be anywhere from 12 to 22 cents on the dollar depending on the actual mix of eliminated positions.
What this analysis does prove: a mid-size American city simultaneously bought AI intake tools and eliminated roughly 19% of its planning department, and the annualized cost of the AI is a fraction of the annualized cost of the labor it accompanied out the door. That pattern will repeat, because the math is irresistible to any budget director looking at a permitting backlog and a headcount line item.
What It Means If You Are Building a House
If you are pulling permits in Denver or any of the dozen-plus cities now deploying AI intake tools, expect faster initial feedback on your application. Expect clearer rejection notices with specific itemized deficiencies instead of vague instructions to "revise and resubmit." Expect the intake process to feel less like mailing a letter into a void.
But also expect that the human reviewer who eventually evaluates your plans for code compliance will be one of fewer reviewers handling a larger share of the workload. Whether "fewer but faster" produces the same quality of plan review as "more but slower" is an empirical question that nobody has studied yet, because the experiment is happening right now in real time on real buildings.
A PermitPlace survey of 741 cities across 44 states found that actual commercial project permitting timelines run two to five times longer than published guidelines, with the published average sitting at 22.9 days and the median at 14, numbers that tell you almost nothing about what you will actually experience on a complex residential project. If you are building something complicated, multiply by three and plan accordingly.
AI might shorten the intake bottleneck, but it will not speed up the code review itself, the structural calculations, or the fire egress analysis that requires a human being who understands what load-bearing means. The risk is that cities use intake automation as justification for staffing levels that make the downstream review slower, not faster, and the applicant ends up trading a shorter wait at the front door for a longer wait in the building.
Denver's bet is that it can do more with less. Maybe. The nine months that Julia Richman lost suggest the old way was broken, but the 59 jobs that disappeared suggest the fix comes with a cost that nobody in the press release wanted to name, and the dozen other cities watching Denver's experiment are already running the same math on their own budgets.